Background: When mapping clinical information to SNOMED CT it is important that mapping
is kept consistent. But selection of SNOMED CT concepts is ambiguous. Therefore, mapping
guidelines are necessary to ensure consistent mapping. Only limited instructions of how to map
clinical information to SNOMED CT are available. [1] have developed a mapping guideline for
mapping EHR-template terms to SNOMED CT. No research studies of how research data should be
mapped to SNOMED CT were found by the literature review conducted in this project.
The objective of this project was to investigate the applicability of [1]s’ mapping guideline for mapping
a research dataset and to investigate how this mapping guideline should be adapted to facilitate
mapping of research data. Further, it was investigated what SNOMED CT may add to a research
dataset which is mapped by use of this mapping guideline.
Method: Investigation was conducted by an exemplification, where a research dataset of 941
syncope patients was mapped to SNOMED CT. The mapping process involved the 3 steps: Grouping
of the DEs of the dataset, mapping each group, and refinements of the mapping. Mapping was
conducted by an iterative mapping process and refinements were conducted until the mapped DEs
fulfilled a set of quality criteria. These quality criteria were used to evaluate the quality of mapping,
thus ensure consistency. Selection of SNOMED CT concepts was conducted according to [1]s’
mapping guideline. For areas which this guideline did not cover mapping was conducted in 3 steps;
1) candidate concepts were identified. 2) An overview of the possibilities and limitations of each
candidate was provided by drawing the defining- and qualifying relationships of each candidate
concept. 3) The best candidate was selected.
Results: The DEs of the dataset was divided into 6 groups. Of these, two groups were selected
("‘diagnoses"’ and "‘medication"’) and mapped to SNOMED CT. It was possible to map the DEs
of each group to subtype descendants of the OE of each group, respectively. Thus, each group
created a subset cluster where the OE was the LCP of the DEs of the group.
Since it was not possible to find appropriate SNOMED CT concepts to represent the contextual data
values of the dataset ("‘Yes"’/"’No"’/"’Unspecified"’) these where not mapped to SNOMED CT. 1 DE
was not mapped, since it was not possible to interpret its semantic meaning.
Conclusion: SNOMED CT provides representation of research data with optional level of
granularity. Further, SNOMED CT adds more details to the dataset. SNOMED CT based research
data is one step towards semantic interoperability and efficient data extraction from EHR-systems,
thus one step towards efficient, high-quality translational research and improved outcome of the
clinical care process.